Untitled
Golnar Babakhani Teimouri
May 19, 2018
Congress tweets after shooting
## # A tibble: 54,333 x 3
## word1 word2 n
## <chr> <chr> <int>
## 1 health care 154
## 2 gun violence 149
## 3 town hall 148
## 4 law enforcement 142
## 5 house floor 129
## 6 op ed 96
## 7 house passed 88
## 8 president obama 82
## 9 navy yard 79
## 10 happy birthday 77
## # ... with 54,323 more rows## # A tibble: 73,798 x 8
## Date Time R_or_D Before_or_After… screen_name hashtags Nominate_dim2
## <chr> <tim> <chr> <chr> <chr> <chr> <dbl>
## 1 1/8/11 00:05 D After GabbyGiffo… <NA> 0.076
## 2 1/8/11 00:05 D After GabbyGiffo… <NA> 0.076
## 3 1/8/11 00:05 D After GabbyGiffo… <NA> 0.076
## 4 1/8/11 00:05 D After GabbyGiffo… <NA> 0.076
## 5 1/8/11 01:53 D After SilvestreR… <NA> 0.304
## 6 1/8/11 01:53 D After SilvestreR… <NA> 0.304
## 7 1/8/11 01:53 D After SilvestreR… <NA> 0.304
## 8 1/8/11 01:53 D After SilvestreR… <NA> 0.304
## 9 1/8/11 01:53 D After SilvestreR… <NA> 0.304
## 10 1/8/11 04:42 D After GabbyGiffo… <NA> 0.076
## # ... with 73,788 more rows, and 1 more variable: bigram <chr>| Republican or Democrat | Term | n | tf | idf | tf_idf |
|---|---|---|---|---|---|
| D | gun violence | 149 | 0.0050074 | 0.6931472 | 0.0034709 |
| D | askdems disarmhate | 36 | 0.0012098 | 0.6931472 | 0.0008386 |
| R | pro life | 40 | 0.0009082 | 0.6931472 | 0.0006295 |
| R | protect children | 39 | 0.0008855 | 0.6931472 | 0.0006138 |
| R | life bill | 37 | 0.0008401 | 0.6931472 | 0.0005823 |
| D | lgbt community | 21 | 0.0007057 | 0.6931472 | 0.0004892 |
| R | obama admin | 30 | 0.0006812 | 0.6931472 | 0.0004721 |
| D | chisummit cbctalks | 20 | 0.0006721 | 0.6931472 | 0.0004659 |
| D | mass shootings | 20 | 0.0006721 | 0.6931472 | 0.0004659 |
| D | earth day | 19 | 0.0006385 | 0.6931472 | 0.0004426 |
| D | minimum wage | 17 | 0.0005713 | 0.6931472 | 0.0003960 |
| R | supports pro | 24 | 0.0005449 | 0.6931472 | 0.0003777 |
| D | senator inouye | 15 | 0.0005041 | 0.6931472 | 0.0003494 |
| R | pa11 nepa | 22 | 0.0004995 | 0.6931472 | 0.0003462 |
| D | altonsterling philandocastile | 14 | 0.0004705 | 0.6931472 | 0.0003261 |
| R | gopoversight hearing | 19 | 0.0004314 | 0.6931472 | 0.0002990 |
| R | 40 hour | 18 | 0.0004087 | 0.6931472 | 0.0002833 |
| R | god bless | 18 | 0.0004087 | 0.6931472 | 0.0002833 |
| R | medicare advantage | 18 | 0.0004087 | 0.6931472 | 0.0002833 |
| R | ne02 omaha | 18 | 0.0004087 | 0.6931472 | 0.0002833 |
Logs Ratio for after shootings
## # A tibble: 54,333 x 4
## bigram D R logratio
## <chr> <dbl> <dbl> <dbl>
## 1 families affected 0.000214 0.000213 -0.00276
## 2 protect americans 0.0000714 0.0000712 -0.00276
## 3 sandy hook 0.0000714 0.0000712 -0.00276
## 4 2 2 0.000202 0.000203 0.00561
## 5 bipartisan bill 0.000202 0.000203 0.00561
## 6 middle school 0.000155 0.000152 -0.0138
## 7 1 2 0.000250 0.000254 0.0174
## 8 care reform 0.0000832 0.0000813 -0.0234
## 9 committee hearing 0.000166 0.000163 -0.0234
## 10 families friends 0.0000832 0.0000813 -0.0234
## # ... with 54,323 more rows
Comparing words frequency by plotting(after shooting)
## # A tibble: 56,844 x 5
## # Groups: R_or_D [2]
## R_or_D bigram n total freq
## <chr> <chr> <int> <int> <dbl>
## 1 D gun violence 149 29756 0.00501
## 2 R law enforcement 103 44042 0.00234
## 3 R health care 96 44042 0.00218
## 4 R house passed 77 44042 0.00175
## 5 R house floor 76 44042 0.00173
## 6 D town hall 75 29756 0.00252
## 7 R town hall 73 44042 0.00166
## 8 R op ed 68 44042 0.00154
## 9 R president obama 62 44042 0.00141
## 10 R watch live 60 44042 0.00136
## # ... with 56,834 more rows## # A tibble: 54,333 x 3
## bigram D R
## <chr> <dbl> <dbl>
## 1 1 15 0.0000336 0.0000227
## 2 1 15pm 0.0000336 0.0000227
## 3 1 40 0.0000336 0.0000227
## 4 1 45 0.0000336 0.0000227
## 5 1.3 billion 0.0000336 0.0000227
## 6 1.9 million 0.0000336 0.0000227
## 7 10 10 0.0000336 0.0000227
## 8 100 million 0.0000336 0.0000227
## 9 100 people 0.0000336 0.0000227
## 10 100 rating 0.0000336 0.0000227
## # ... with 54,323 more rows
Congress tweets before shootings
## # A tibble: 104,813 x 2
## word n
## <fct> <int>
## 1 of the 719
## 2 on the 555
## 3 in the 524
## 4 to the 494
## 5 the house 333
## 6 for the 327
## 7 at the 275
## 8 thank you 236
## 9 will be 216
## 10 proud to 207
## # ... with 104,803 more rows## # A tibble: 38,515 x 3
## word1 word2 n
## <chr> <chr> <int>
## 1 health care 111
## 2 house floor 111
## 3 town hall 78
## 4 watch live 74
## 5 op ed 72
## 6 middle class 70
## 7 american people 63
## 8 national security 59
## 9 happy birthday 56
## 10 president obama 47
## # ... with 38,505 more rows## # A tibble: 49,640 x 12
## Date Time R_or_D Before_or_After… screen_name hashtags Nominate_dim2
## <chr> <tim> <chr> <chr> <chr> <chr> <dbl>
## 1 1/10/… 00:02 D Before RepVisclos… <NA> 0.246
## 2 1/10/… 00:02 D Before RepVisclos… <NA> 0.246
## 3 1/10/… 00:02 D Before RepVisclos… <NA> 0.246
## 4 1/10/… 00:02 D Before RepVisclos… <NA> 0.246
## 5 1/10/… 00:59 R Before RepKayGran… <NA> 0.231
## 6 1/10/… 00:59 R Before RepKayGran… <NA> 0.231
## 7 1/10/… 00:59 R Before RepKayGran… <NA> 0.231
## 8 1/10/… 03:39 R Before DevinNunes <NA> 0.194
## 9 1/10/… 03:39 R Before DevinNunes <NA> 0.194
## 10 1/10/… 03:39 R Before DevinNunes <NA> 0.194
## # ... with 49,630 more rows, and 5 more variables: X9 <chr>, X10 <chr>,
## # X11 <chr>, X12 <chr>, bigram <chr>| Republican or Democrat | Term | n | tf | idf | tf_idf |
|---|---|---|---|---|---|
| D | gun violence | 26 | 0.0013760 | 0.6931472 | 0.0009538 |
| D | senator inouye | 15 | 0.0007939 | 0.6931472 | 0.0005503 |
| R | veterans deserve | 24 | 0.0007806 | 0.6931472 | 0.0005411 |
| D | equal pay | 14 | 0.0007409 | 0.6931472 | 0.0005136 |
| D | class tax | 12 | 0.0006351 | 0.6931472 | 0.0004402 |
| D | class security | 11 | 0.0005822 | 0.6931472 | 0.0004035 |
| R | tax reform | 17 | 0.0005529 | 0.6931472 | 0.0003833 |
| D | americans makeitinamerica | 10 | 0.0005292 | 0.6931472 | 0.0003668 |
| D | disarmhate lightingtheway | 10 | 0.0005292 | 0.6931472 | 0.0003668 |
| D | student debt | 10 | 0.0005292 | 0.6931472 | 0.0003668 |
| R | floor speech | 16 | 0.0005204 | 0.6931472 | 0.0003607 |
| D | 9 hawaii | 8 | 0.0004234 | 0.6931472 | 0.0002935 |
| D | critical support | 8 | 0.0004234 | 0.6931472 | 0.0002935 |
| D | dotherightthing extend | 8 | 0.0004234 | 0.6931472 | 0.0002935 |
| D | innovation job | 8 | 0.0004234 | 0.6931472 | 0.0002935 |
| D | paycheck fairness | 8 | 0.0004234 | 0.6931472 | 0.0002935 |
| R | sen collins | 13 | 0.0004228 | 0.6931472 | 0.0002931 |
| R | care bill | 12 | 0.0003903 | 0.6931472 | 0.0002705 |
| R | sanctions relief | 12 | 0.0003903 | 0.6931472 | 0.0002705 |
| R | tax hike | 12 | 0.0003903 | 0.6931472 | 0.0002705 |
#Logs Ratio for before shooting
## # A tibble: 38,515 x 4
## bigram D R logratio
## <chr> <dbl> <dbl> <dbl>
## 1 1 million 0.0000871 0.0000866 -0.00533
## 2 7 30 0.0000871 0.0000866 -0.00533
## 3 save lives 0.0000871 0.0000866 -0.00533
## 4 step forward 0.0000871 0.0000866 -0.00533
## 5 30 pm 0.000105 0.000101 -0.0335
## 6 american flag 0.000105 0.000101 -0.0335
## 7 gabrielle giffords 0.000105 0.000101 -0.0335
## 8 6 30 0.0000697 0.0000722 0.0355
## 9 appropriations act 0.0000697 0.0000722 0.0355
## 10 beautiful day 0.0000697 0.0000722 0.0355
## # ... with 38,505 more rows
#Comparing all graphs
Joining before and after shooting dataframes
## # A tibble: 87,300 x 3
## word1 word2 n
## <chr> <chr> <int>
## 1 health care 265
## 2 house floor 240
## 3 town hall 226
## 4 gun violence 175
## 5 op ed 168
## 6 law enforcement 161
## 7 watch live 144
## 8 happy birthday 133
## 9 president obama 129
## 10 house passed 126
## # ... with 87,290 more rows## # A tibble: 123,413 x 12
## Date Time R_or_D Before_or_After… screen_name hashtags Nominate_dim2
## <chr> <tim> <chr> <chr> <chr> <chr> <dbl>
## 1 1/10/… 00:02 D Before RepVisclos… <NA> 0.246
## 2 1/10/… 00:02 D Before RepVisclos… <NA> 0.246
## 3 1/10/… 00:02 D Before RepVisclos… <NA> 0.246
## 4 1/10/… 00:02 D Before RepVisclos… <NA> 0.246
## 5 1/10/… 00:59 R Before RepKayGran… <NA> 0.231
## 6 1/10/… 00:59 R Before RepKayGran… <NA> 0.231
## 7 1/10/… 00:59 R Before RepKayGran… <NA> 0.231
## 8 1/10/… 03:39 R Before DevinNunes <NA> 0.194
## 9 1/10/… 03:39 R Before DevinNunes <NA> 0.194
## 10 1/10/… 03:39 R Before DevinNunes <NA> 0.194
## # ... with 123,403 more rows, and 5 more variables: X9 <chr>, X10 <chr>,
## # X11 <chr>, X12 <chr>, bigram <chr>Words over time
## # A tibble: 910 x 6
## time_floor R_or_D bigram count time_total word_total
## <dttm> <chr> <chr> <int> <int> <int>
## 1 2009-01-01 00:00:00 D 30 pm 1 596 43
## 2 2009-01-01 00:00:00 D american people 1 596 120
## 3 2009-01-01 00:00:00 D capitol hill 1 596 45
## 4 2009-01-01 00:00:00 D committee heari… 1 596 36
## 5 2009-01-01 00:00:00 D fort hood 1 596 63
## 6 2009-01-01 00:00:00 D health care 7 596 265
## 7 2009-01-01 00:00:00 D health insurance 2 596 41
## 8 2009-01-01 00:00:00 D house floor 2 596 240
## 9 2009-01-01 00:00:00 D listen live 4 596 72
## 10 2009-01-01 00:00:00 D op ed 2 596 168
## # ... with 900 more rows## # A tibble: 200 x 3
## R_or_D bigram data
## <chr> <chr> <list>
## 1 D 30 pm <tibble [6 × 4]>
## 2 D american people <tibble [6 × 4]>
## 3 D capitol hill <tibble [5 × 4]>
## 4 D committee hearing <tibble [7 × 4]>
## 5 D fort hood <tibble [2 × 4]>
## 6 D health care <tibble [7 × 4]>
## 7 D health insurance <tibble [7 × 4]>
## 8 D house floor <tibble [7 × 4]>
## 9 D listen live <tibble [5 × 4]>
## 10 D op ed <tibble [7 × 4]>
## # ... with 190 more rows## # A tibble: 200 x 4
## R_or_D bigram data models
## <chr> <chr> <list> <list>
## 1 D 30 pm <tibble [6 × 4]> <S3: glm>
## 2 D american people <tibble [6 × 4]> <S3: glm>
## 3 D capitol hill <tibble [5 × 4]> <S3: glm>
## 4 D committee hearing <tibble [7 × 4]> <S3: glm>
## 5 D fort hood <tibble [2 × 4]> <S3: glm>
## 6 D health care <tibble [7 × 4]> <S3: glm>
## 7 D health insurance <tibble [7 × 4]> <S3: glm>
## 8 D house floor <tibble [7 × 4]> <S3: glm>
## 9 D listen live <tibble [5 × 4]> <S3: glm>
## 10 D op ed <tibble [7 × 4]> <S3: glm>
## # ... with 190 more rows## # A tibble: 24 x 8
## R_or_D bigram term estimate std.error statistic p.value
## <chr> <chr> <chr> <dbl> <dbl> <dbl> <dbl>
## 1 D health care time_flo… -7.07e-9 1.75e-9 -4.05 5.21e- 5
## 2 D listen live time_flo… -2.20e-8 5.11e-9 -4.31 1.66e- 5
## 3 D police office… time_flo… 2.07e-8 4.87e-9 4.25 2.11e- 5
## 4 R 30 pm time_flo… -1.44e-8 2.58e-9 -5.57 2.56e- 8
## 5 R fort hood time_flo… -1.14e-8 1.81e-9 -6.27 3.61e-10
## 6 R health care time_flo… -1.31e-8 1.20e-9 -11.0 4.22e-28
## 7 R house gop time_flo… -1.72e-8 2.85e-9 -6.06 1.36e- 9
## 8 R press confere… time_flo… -1.12e-8 2.65e-9 -4.22 2.46e- 5
## 9 R town hall time_flo… -6.14e-9 1.51e-9 -4.06 4.96e- 5
## 10 R create jobs time_flo… -1.53e-8 2.69e-9 -5.70 1.22e- 8
## # ... with 14 more rows, and 1 more variable: adjusted.p.value <dbl>
Building a dtm to find association
## <<SimpleCorpus>>
## Metadata: corpus specific: 1, document level (indexed): 0
## Content: documents: 17902## <<TermDocumentMatrix (terms: 19813, documents: 17902)>>
## Non-/sparse entries: 184359/354507967
## Sparsity : 100%
## Maximal term length: 45
## Weighting : term frequency (tf)## $terror
## iranian notonec iran sponsor
## 0.34 0.30 0.21 0.20## <<SimpleCorpus>>
## Metadata: corpus specific: 1, document level (indexed): 0
## Content: documents: 11680## <<TermDocumentMatrix (terms: 15548, documents: 11680)>>
## Non-/sparse entries: 119964/181480676
## Sparsity : 100%
## Maximal term length: 38
## Weighting : term frequency (tf)## $terror
## notonec sponsor franc nice
## 0.24 0.22 0.21 0.20## <<SimpleCorpus>>
## Metadata: corpus specific: 1, document level (indexed): 0
## Content: documents: 10718## <<TermDocumentMatrix (terms: 13466, documents: 10718)>>
## Non-/sparse entries: 108303/144220285
## Sparsity : 100%
## Maximal term length: 32
## Weighting : term frequency (tf)
|
||||||||||||||||||
| 1 Footnote1; notonec refers to the hashtag NotOneCent. The hashtag was used during Iran nuclear negotions by those who opposed unfreezing Iran’s assetts. |
## <<SimpleCorpus>>
## Metadata: corpus specific: 1, document level (indexed): 0
## Content: documents: 7184## <<TermDocumentMatrix (terms: 11467, documents: 7184)>>
## Non-/sparse entries: 76056/82302872
## Sparsity : 100%
## Maximal term length: 45
## Weighting : term frequency (tf)
|
## <<SimpleCorpus>>
## Metadata: corpus specific: 1, document level (indexed): 0
## Content: documents: 7342## <<TermDocumentMatrix (terms: 11027, documents: 7342)>>
## Non-/sparse entries: 74195/80886039
## Sparsity : 100%
## Maximal term length: 33
## Weighting : term frequency (tf)
|
||||||||||
| 1 Footnote1; franc refers to terrorist attack in Nice, France. The attack occured on July 14,2016 exactly 48 hours before the Bouton Rouge Mass Shooting in the U.S. | ||||||||||
| 2 Footnote2; notonec refers to the hashtag NotOneCent. The hashtag was used during Iran nuclear negotions by those who opposed unfreezing Iran’s assetts. |
## <<SimpleCorpus>>
## Metadata: corpus specific: 1, document level (indexed): 0
## Content: documents: 4337## <<TermDocumentMatrix (terms: 8613, documents: 4337)>>
## Non-/sparse entries: 45769/37308812
## Sparsity : 100%
## Maximal term length: 38
## Weighting : term frequency (tf)
|
||||||||||||||||||||||||||||||||||||||||
| 1 Footnote1; nice refers to terrorist attack in Nice, France. The attack occured on July 14,2016 exactly 48 hours before the Bouton Rouge Mass Shooting in the U.S. |
Sentiment Analysis
Need my tokens to be words instead of bigrams for sentiment analysis. So i change my tokens to words first
Logistic Regression
## # A tibble: 475 x 16
## Date Time R_or_D Before_or_After… screen_name text hashtags
## <chr> <time> <chr> <chr> <chr> <chr> <chr>
## 1 12/3/… 13:10 R After Robert_Ade… Contortions… SanBern…
## 2 6/12/… 19:02 R After RepMarthaR… We must dis… <NA>
## 3 6/12/… 18:59 R After RepMarthaR… This is the… <NA>
## 4 6/12/… 18:51 R After RepMarthaR… Horrified &… <NA>
## 5 6/12/… 18:24 R After USRepGaryP… My thoughts… Orlando
## 6 10/1/… 19:02 R After USRepGaryP… "#NotOneCen… NotOneC…
## 7 10/1/… 16:10 R After USRepGaryP… See @RepMee… NotOneC…
## 8 6/11/… 22:16 R After USRepGaryP… "Proud to s… <NA>
## 9 10/2/… 15:02 R After RepFrenchH… ICYMI: My s… <NA>
## 10 10/1/… 21:23 R After RepFrenchH… Here is why… NotOneC…
## # ... with 465 more rows, and 9 more variables: Nominate_dim2 <dbl>,
## # X9 <chr>, X10 <chr>, X11 <chr>, X12 <chr>, case <chr>,
## # total_victims <int>, mental_health_issues <chr>, race <chr>
##
## Call:
## glm(formula = mentions_terror ~ R_or_D + Nominate_dim2, family = "binomial",
## data = df2)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -0.2107 -0.1998 -0.1959 -0.1164 3.1945
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -5.0111 0.1151 -43.525 <2e-16 ***
## R_or_DR 1.1076 0.1279 8.659 <2e-16 ***
## Nominate_dim2 -0.1411 0.1842 -0.766 0.444
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 4509.3 on 29579 degrees of freedom
## Residual deviance: 4416.9 on 29577 degrees of freedom
## (45 observations deleted due to missingness)
## AIC: 4422.9
##
## Number of Fisher Scoring iterations: 7## # A tibble: 692 x 3
## R_or_D Nominate_dim2 pred
## <fct> <dbl> <dbl>
## 1 D -0.758 -4.90
## 2 D -0.752 -4.91
## 3 D -0.696 -4.91
## 4 D -0.659 -4.92
## 5 D -0.647 -4.92
## 6 D -0.62 -4.92
## 7 D -0.602 -4.93
## 8 D -0.571 -4.93
## 9 D -0.563 -4.93
## 10 D -0.546 -4.93
## # ... with 682 more rows
##
## Call:
## glm(formula = Shooting_related_terrorism ~ R_or_D + Nominate_dim2 +
## race + race * R_or_D + race * Nominate_dim2, family = binomial(link = "logit"),
## data = df3)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -8.49 0.00 0.00 0.00 8.49
##
## Coefficients: (1 not defined because of singularities)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -4.172e+15 8.542e+07 -4.884e+07 <2e-16 ***
## R_or_DR 4.172e+15 7.787e+07 5.358e+07 <2e-16 ***
## Nominate_dim2 4.952e+10 1.002e+08 4.939e+02 <2e-16 ***
## raceOther 6.015e+15 8.636e+07 6.966e+07 <2e-16 ***
## raceWhite 3.690e+15 7.804e+07 4.728e+07 <2e-16 ***
## R_or_DR:raceOther -3.433e+15 7.932e+07 -4.329e+07 <2e-16 ***
## R_or_DR:raceWhite NA NA NA NA
## Nominate_dim2:raceOther -2.736e+15 1.035e+08 -2.644e+07 <2e-16 ***
## Nominate_dim2:raceWhite 4.625e+15 1.466e+08 3.154e+07 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 53.00 on 116 degrees of freedom
## Residual deviance: 504.61 on 109 degrees of freedom
## (358 observations deleted due to missingness)
## AIC: 520.61
##
## Number of Fisher Scoring iterations: 24## Likelihood ratio test
##
## Model 1: Shooting_related_terrorism ~ R_or_D + Nominate_dim2 + race +
## race * R_or_D + race * Nominate_dim2
## Model 2: Shooting_related_terrorism ~ R_or_D + Nominate_dim2 + race
## #Df LogLik Df Chisq Pr(>Chisq)
## 1 7 -7.0747
## 2 5 -7.0747 -2 0 1## llh llhNull G2 McFadden r2ML
## -7.0747493 -133.2836519 252.4178053 0.9469196 0.9898412
## r2CU
## 0.9976774## # A tibble: 1,272 x 4
## R_or_D Nominate_dim2 race pred
## <fct> <dbl> <fct> <dbl>
## 1 D -0.647 Latino -4.17e15
## 2 D -0.647 Other 3.61e15
## 3 D -0.647 White -3.47e15
## 4 D -0.647 <NA> NA
## 5 D -0.563 Latino -4.17e15
## 6 D -0.563 Other 3.38e15
## 7 D -0.563 White -3.09e15
## 8 D -0.563 <NA> NA
## 9 D -0.492 Latino -4.17e15
## 10 D -0.492 Other 3.19e15
## # ... with 1,262 more rows## Classes 'tbl_df', 'tbl' and 'data.frame': 954 obs. of 4 variables:
## $ R_or_D : Factor w/ 2 levels "D","R": 1 1 1 1 1 1 1 1 1 1 ...
## $ Nominate_dim2: num -0.647 -0.647 -0.647 -0.563 -0.563 -0.563 -0.492 -0.492 -0.492 -0.49 ...
## $ race : Factor w/ 3 levels "Latino","Other",..: 1 2 3 1 2 3 1 2 3 1 ...
## $ pred : num -4.17e+15 3.61e+15 -3.47e+15 -4.17e+15 3.38e+15 ...